Towards an automated nonparametric computerized clinical electrocardiograph

Mustafa, A. M.; Mismar, M. J.

A large extent of the information contained in time-domain ECG is not utilized. The full utilization or this information is largely based on the experience and personal judgment. However, the existence of relatively cheap, multipurpose reliable digital computers makes it possible, to furnish clinics with appropriate hardware, associated software and algorithms that make these computers automatic ECG identifiers. This study is involved with the development of functions and services, usually offered by a clinical computer to include facilities for ECG signals, amplifications, digitization, filtration, storage, signal processing, information extraction, and abnormality identification. The user, with the aid of menues, is allowed to select, stretch, compress and search data for suspected abnormalities. In addition, essential parameters usually derived from ECGs such as the heart rate, the R-R wave and the P-Q intervals,.. etc. are evaluated and forecasted on the computer monitor. The FFT and the Nearest Neighbor algorithms are employed to create a compressed frequency domain database and develop a report on the identity of a possible abnormality, with 1 certain probability of error. This is demonstrated on easy to read and evaluate classification tables. This automation process can be invested for the advantage of the patient and physician in critical times.